]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blame - init/main.c
init/main.c: remove unused declaration of tc_init()
[mirror_ubuntu-eoan-kernel.git] / init / main.c
CommitLineData
1da177e4
LT
1/*
2 * linux/init/main.c
3 *
4 * Copyright (C) 1991, 1992 Linus Torvalds
5 *
6 * GK 2/5/95 - Changed to support mounting root fs via NFS
7 * Added initrd & change_root: Werner Almesberger & Hans Lermen, Feb '96
8 * Moan early if gcc is old, avoiding bogus kernels - Paul Gortmaker, May '96
9 * Simplified starting of init: Michael A. Griffith <grif@acm.org>
10 */
11
ea676e84
AM
12#define DEBUG /* Enable initcall_debug */
13
1da177e4
LT
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/proc_fs.h>
1da177e4
LT
17#include <linux/kernel.h>
18#include <linux/syscalls.h>
9b5609fd 19#include <linux/stackprotector.h>
1da177e4
LT
20#include <linux/string.h>
21#include <linux/ctype.h>
22#include <linux/delay.h>
1da177e4
LT
23#include <linux/ioport.h>
24#include <linux/init.h>
1da177e4 25#include <linux/initrd.h>
1da177e4 26#include <linux/bootmem.h>
4a7a16dc 27#include <linux/acpi.h>
1da177e4 28#include <linux/tty.h>
1da177e4
LT
29#include <linux/percpu.h>
30#include <linux/kmod.h>
db64fe02 31#include <linux/vmalloc.h>
1da177e4 32#include <linux/kernel_stat.h>
d7cd5611 33#include <linux/start_kernel.h>
1da177e4 34#include <linux/security.h>
3d442233 35#include <linux/smp.h>
1da177e4
LT
36#include <linux/profile.h>
37#include <linux/rcupdate.h>
38#include <linux/moduleparam.h>
39#include <linux/kallsyms.h>
40#include <linux/writeback.h>
41#include <linux/cpu.h>
42#include <linux/cpuset.h>
ddbcc7e8 43#include <linux/cgroup.h>
1da177e4 44#include <linux/efi.h>
906568c9 45#include <linux/tick.h>
6168a702 46#include <linux/interrupt.h>
c757249a 47#include <linux/taskstats_kern.h>
ca74e92b 48#include <linux/delayacct.h>
1da177e4
LT
49#include <linux/unistd.h>
50#include <linux/rmap.h>
51#include <linux/mempolicy.h>
52#include <linux/key.h>
b6cd0b77 53#include <linux/buffer_head.h>
94b6da5a 54#include <linux/page_cgroup.h>
9a11b49a 55#include <linux/debug_locks.h>
3ac7fe5a 56#include <linux/debugobjects.h>
fbb9ce95 57#include <linux/lockdep.h>
3c7b4e6b 58#include <linux/kmemleak.h>
84d73786 59#include <linux/pid_namespace.h>
1f21782e 60#include <linux/device.h>
73c27992 61#include <linux/kthread.h>
e6fe6649 62#include <linux/sched.h>
a1c9eea9 63#include <linux/signal.h>
199f0ca5 64#include <linux/idr.h>
0b4b3827 65#include <linux/kgdb.h>
68bf21aa 66#include <linux/ftrace.h>
22a9d645 67#include <linux/async.h>
dfec072e 68#include <linux/kmemcheck.h>
6ae6996a 69#include <linux/sfi.h>
2b2af54a 70#include <linux/shmem_fs.h>
5a0e3ad6 71#include <linux/slab.h>
24a24bb6 72#include <linux/perf_event.h>
4a9d4b02 73#include <linux/file.h>
a74fb73c 74#include <linux/ptrace.h>
bb813f4c
TH
75#include <linux/blkdev.h>
76#include <linux/elevator.h>
38ff87f7 77#include <linux/sched_clock.h>
65f382fd 78#include <linux/context_tracking.h>
47d06e53 79#include <linux/random.h>
1da177e4
LT
80
81#include <asm/io.h>
82#include <asm/bugs.h>
83#include <asm/setup.h>
a940199f 84#include <asm/sections.h>
37b73c82 85#include <asm/cacheflush.h>
1da177e4 86
1da177e4
LT
87#ifdef CONFIG_X86_LOCAL_APIC
88#include <asm/smp.h>
89#endif
90
aae5f662 91static int kernel_init(void *);
1da177e4
LT
92
93extern void init_IRQ(void);
1da177e4
LT
94extern void fork_init(unsigned long);
95extern void mca_init(void);
96extern void sbus_init(void);
1da177e4 97extern void radix_tree_init(void);
37b73c82
AV
98#ifndef CONFIG_DEBUG_RODATA
99static inline void mark_rodata_ro(void) { }
100#endif
1da177e4 101
2ce802f6
TH
102/*
103 * Debug helper: via this flag we know that we are in 'early bootup code'
104 * where only the boot processor is running with IRQ disabled. This means
105 * two things - IRQ must not be enabled before the flag is cleared and some
106 * operations which are not allowed with IRQ disabled are allowed while the
107 * flag is set.
108 */
109bool early_boot_irqs_disabled __read_mostly;
110
a6826048 111enum system_states system_state __read_mostly;
1da177e4
LT
112EXPORT_SYMBOL(system_state);
113
114/*
115 * Boot command-line arguments
116 */
117#define MAX_INIT_ARGS CONFIG_INIT_ENV_ARG_LIMIT
118#define MAX_INIT_ENVS CONFIG_INIT_ENV_ARG_LIMIT
119
120extern void time_init(void);
121/* Default late time init is NULL. archs can override this later. */
d2e3192b 122void (*__initdata late_time_init)(void);
1da177e4 123
30d7e0d4
ABL
124/* Untouched command line saved by arch-specific code. */
125char __initdata boot_command_line[COMMAND_LINE_SIZE];
126/* Untouched saved command line (eg. for /proc) */
127char *saved_command_line;
128/* Command line for parameter parsing */
129static char *static_command_line;
08746a65
KM
130/* Command line for per-initcall parameter parsing */
131static char *initcall_command_line;
1da177e4
LT
132
133static char *execute_command;
ffdfc409 134static char *ramdisk_execute_command;
1da177e4 135
c4b2c0c5
HFS
136/*
137 * Used to generate warnings if static_key manipulation functions are used
138 * before jump_label_init is called.
139 */
140bool static_key_initialized __read_mostly = false;
141EXPORT_SYMBOL_GPL(static_key_initialized);
142
8b3b2955
JB
143/*
144 * If set, this is an indication to the drivers that reset the underlying
145 * device before going ahead with the initialization otherwise driver might
146 * rely on the BIOS and skip the reset operation.
147 *
148 * This is useful if kernel is booting in an unreliable environment.
149 * For ex. kdump situaiton where previous kernel has crashed, BIOS has been
150 * skipped and devices will be in unknown state.
151 */
152unsigned int reset_devices;
153EXPORT_SYMBOL(reset_devices);
1da177e4 154
7e96287d
VG
155static int __init set_reset_devices(char *str)
156{
157 reset_devices = 1;
158 return 1;
159}
160
161__setup("reset_devices", set_reset_devices);
162
d7627467
DH
163static const char * argv_init[MAX_INIT_ARGS+2] = { "init", NULL, };
164const char * envp_init[MAX_INIT_ENVS+2] = { "HOME=/", "TERM=linux", NULL, };
1da177e4
LT
165static const char *panic_later, *panic_param;
166
914dcaa8 167extern const struct obs_kernel_param __setup_start[], __setup_end[];
1da177e4
LT
168
169static int __init obsolete_checksetup(char *line)
170{
914dcaa8 171 const struct obs_kernel_param *p;
33df0d19 172 int had_early_param = 0;
1da177e4
LT
173
174 p = __setup_start;
175 do {
176 int n = strlen(p->str);
b1e4d20c 177 if (parameqn(line, p->str, n)) {
1da177e4 178 if (p->early) {
33df0d19
RR
179 /* Already done in parse_early_param?
180 * (Needs exact match on param part).
181 * Keep iterating, as we can have early
182 * params and __setups of same names 8( */
1da177e4 183 if (line[n] == '\0' || line[n] == '=')
33df0d19 184 had_early_param = 1;
1da177e4 185 } else if (!p->setup_func) {
ea676e84
AM
186 pr_warn("Parameter %s is obsolete, ignored\n",
187 p->str);
1da177e4
LT
188 return 1;
189 } else if (p->setup_func(line + n))
190 return 1;
191 }
192 p++;
193 } while (p < __setup_end);
33df0d19
RR
194
195 return had_early_param;
1da177e4
LT
196}
197
198/*
199 * This should be approx 2 Bo*oMips to start (note initial shift), and will
200 * still work even if initially too large, it will just take slightly longer
201 */
202unsigned long loops_per_jiffy = (1<<12);
203
204EXPORT_SYMBOL(loops_per_jiffy);
205
206static int __init debug_kernel(char *str)
207{
1da177e4 208 console_loglevel = 10;
f6f21c81 209 return 0;
1da177e4
LT
210}
211
212static int __init quiet_kernel(char *str)
213{
1da177e4 214 console_loglevel = 4;
f6f21c81 215 return 0;
1da177e4
LT
216}
217
f6f21c81
YL
218early_param("debug", debug_kernel);
219early_param("quiet", quiet_kernel);
1da177e4
LT
220
221static int __init loglevel(char *str)
222{
808bf29b
AS
223 int newlevel;
224
225 /*
226 * Only update loglevel value when a correct setting was passed,
227 * to prevent blind crashes (when loglevel being set to 0) that
228 * are quite hard to debug
229 */
230 if (get_option(&str, &newlevel)) {
231 console_loglevel = newlevel;
232 return 0;
233 }
234
235 return -EINVAL;
1da177e4
LT
236}
237
f6f21c81 238early_param("loglevel", loglevel);
1da177e4 239
a99cd112 240/* Change NUL term back to "=", to make "param" the whole string. */
eb157427 241static int __init repair_env_string(char *param, char *val, const char *unused)
1da177e4 242{
1da177e4
LT
243 if (val) {
244 /* param=val or param="val"? */
245 if (val == param+strlen(param)+1)
246 val[-1] = '=';
247 else if (val == param+strlen(param)+2) {
248 val[-2] = '=';
249 memmove(val-1, val, strlen(val)+1);
250 val--;
251 } else
252 BUG();
253 }
a99cd112
CM
254 return 0;
255}
256
257/*
258 * Unknown boot options get handed to init, unless they look like
259 * unused parameters (modprobe will find them in /proc/cmdline).
260 */
eb157427 261static int __init unknown_bootoption(char *param, char *val, const char *unused)
a99cd112 262{
eb157427 263 repair_env_string(param, val, unused);
1da177e4
LT
264
265 /* Handle obsolete-style parameters */
266 if (obsolete_checksetup(param))
267 return 0;
268
f066a4f6
RR
269 /* Unused module parameter. */
270 if (strchr(param, '.') && (!val || strchr(param, '.') < val))
1da177e4 271 return 0;
1da177e4
LT
272
273 if (panic_later)
274 return 0;
275
276 if (val) {
277 /* Environment option */
278 unsigned int i;
279 for (i = 0; envp_init[i]; i++) {
280 if (i == MAX_INIT_ENVS) {
281 panic_later = "Too many boot env vars at `%s'";
282 panic_param = param;
283 }
284 if (!strncmp(param, envp_init[i], val - param))
285 break;
286 }
287 envp_init[i] = param;
288 } else {
289 /* Command line option */
290 unsigned int i;
291 for (i = 0; argv_init[i]; i++) {
292 if (i == MAX_INIT_ARGS) {
293 panic_later = "Too many boot init vars at `%s'";
294 panic_param = param;
295 }
296 }
297 argv_init[i] = param;
298 }
299 return 0;
300}
301
302static int __init init_setup(char *str)
303{
304 unsigned int i;
305
306 execute_command = str;
307 /*
308 * In case LILO is going to boot us with default command line,
309 * it prepends "auto" before the whole cmdline which makes
310 * the shell think it should execute a script with such name.
311 * So we ignore all arguments entered _before_ init=... [MJ]
312 */
313 for (i = 1; i < MAX_INIT_ARGS; i++)
314 argv_init[i] = NULL;
315 return 1;
316}
317__setup("init=", init_setup);
318
ffdfc409
OJ
319static int __init rdinit_setup(char *str)
320{
321 unsigned int i;
322
323 ramdisk_execute_command = str;
324 /* See "auto" comment in init_setup */
325 for (i = 1; i < MAX_INIT_ARGS; i++)
326 argv_init[i] = NULL;
327 return 1;
328}
329__setup("rdinit=", rdinit_setup);
330
1da177e4 331#ifndef CONFIG_SMP
34db18a0 332static const unsigned int setup_max_cpus = NR_CPUS;
1da177e4
LT
333#ifdef CONFIG_X86_LOCAL_APIC
334static void __init smp_init(void)
335{
336 APIC_init_uniprocessor();
337}
338#else
339#define smp_init() do { } while (0)
340#endif
341
e0982e90 342static inline void setup_nr_cpu_ids(void) { }
1da177e4 343static inline void smp_prepare_cpus(unsigned int maxcpus) { }
1da177e4
LT
344#endif
345
30d7e0d4
ABL
346/*
347 * We need to store the untouched command line for future reference.
348 * We also need to store the touched command line since the parameter
349 * parsing is performed in place, and we should allow a component to
350 * store reference of name/value for future reference.
351 */
352static void __init setup_command_line(char *command_line)
353{
098b081b
SS
354 saved_command_line =
355 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
356 initcall_command_line =
357 memblock_virt_alloc(strlen(boot_command_line) + 1, 0);
358 static_command_line = memblock_virt_alloc(strlen(command_line) + 1, 0);
30d7e0d4
ABL
359 strcpy (saved_command_line, boot_command_line);
360 strcpy (static_command_line, command_line);
361}
362
1da177e4
LT
363/*
364 * We need to finalize in a non-__init function or else race conditions
365 * between the root thread and the init thread may cause start_kernel to
366 * be reaped by free_initmem before the root thread has proceeded to
367 * cpu_idle.
368 *
369 * gcc-3.4 accidentally inlines this function, so use noinline.
370 */
371
b433c3d4
PZ
372static __initdata DECLARE_COMPLETION(kthreadd_done);
373
f99ebf0a 374static noinline void __init_refok rest_init(void)
1da177e4 375{
73c27992
EB
376 int pid;
377
7db905e6 378 rcu_scheduler_starting();
b433c3d4 379 /*
97158569 380 * We need to spawn init first so that it obtains pid 1, however
b433c3d4
PZ
381 * the init task will end up wanting to create kthreads, which, if
382 * we schedule it before we create kthreadd, will OOPS.
383 */
aae5f662 384 kernel_thread(kernel_init, NULL, CLONE_FS | CLONE_SIGHAND);
1da177e4 385 numa_default_policy();
73c27992 386 pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);
d11c563d 387 rcu_read_lock();
5cd20455 388 kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);
d11c563d 389 rcu_read_unlock();
b433c3d4 390 complete(&kthreadd_done);
f340c0d1
IM
391
392 /*
393 * The boot idle thread must execute schedule()
1df21055 394 * at least once to get things moving:
f340c0d1 395 */
1df21055 396 init_idle_bootup_task(current);
bd2f5536 397 schedule_preempt_disabled();
5bfb5d69 398 /* Call into cpu_idle with preempt disabled */
a1a04ec3 399 cpu_startup_entry(CPUHP_ONLINE);
1df21055 400}
1da177e4
LT
401
402/* Check for early params. */
9fb48c74 403static int __init do_early_param(char *param, char *val, const char *unused)
1da177e4 404{
914dcaa8 405 const struct obs_kernel_param *p;
1da177e4
LT
406
407 for (p = __setup_start; p < __setup_end; p++) {
b1e4d20c 408 if ((p->early && parameq(param, p->str)) ||
18a8bd94
YL
409 (strcmp(param, "console") == 0 &&
410 strcmp(p->str, "earlycon") == 0)
411 ) {
1da177e4 412 if (p->setup_func(val) != 0)
ea676e84 413 pr_warn("Malformed early option '%s'\n", param);
1da177e4
LT
414 }
415 }
416 /* We accept everything at this stage. */
417 return 0;
418}
419
13977091
MD
420void __init parse_early_options(char *cmdline)
421{
026cee00 422 parse_args("early options", cmdline, NULL, 0, 0, 0, do_early_param);
13977091
MD
423}
424
1da177e4
LT
425/* Arch code calls this early on, or if not, just before other parsing. */
426void __init parse_early_param(void)
427{
428 static __initdata int done = 0;
429 static __initdata char tmp_cmdline[COMMAND_LINE_SIZE];
430
431 if (done)
432 return;
433
434 /* All fall through to do_early_param. */
30d7e0d4 435 strlcpy(tmp_cmdline, boot_command_line, COMMAND_LINE_SIZE);
13977091 436 parse_early_options(tmp_cmdline);
1da177e4
LT
437 done = 1;
438}
439
440/*
441 * Activate the first processor.
442 */
443
44fd2299
SS
444static void __init boot_cpu_init(void)
445{
446 int cpu = smp_processor_id();
447 /* Mark the boot cpu "present", "online" etc for SMP and UP case */
915441b6 448 set_cpu_online(cpu, true);
933b0618 449 set_cpu_active(cpu, true);
915441b6
RR
450 set_cpu_present(cpu, true);
451 set_cpu_possible(cpu, true);
44fd2299
SS
452}
453
839ad62e 454void __init __weak smp_setup_processor_id(void)
033ab7f8
AM
455{
456}
457
eded09cc 458# if THREAD_SIZE >= PAGE_SIZE
8c9843e5
BH
459void __init __weak thread_info_cache_init(void)
460{
461}
eded09cc 462#endif
8c9843e5 463
444f478f
PE
464/*
465 * Set up kernel memory allocators
466 */
467static void __init mm_init(void)
468{
ca371c0d 469 /*
7fa87ce7
JC
470 * page_cgroup requires contiguous pages,
471 * bigger than MAX_ORDER unless SPARSEMEM.
ca371c0d
KH
472 */
473 page_cgroup_init_flatmem();
444f478f
PE
474 mem_init();
475 kmem_cache_init();
099a19d9 476 percpu_init_late();
b35f1819 477 pgtable_init();
444f478f
PE
478 vmalloc_init();
479}
480
1da177e4
LT
481asmlinkage void __init start_kernel(void)
482{
483 char * command_line;
914dcaa8 484 extern const struct kernel_param __start___param[], __stop___param[];
033ab7f8 485
fbb9ce95
IM
486 /*
487 * Need to run as early as possible, to initialize the
488 * lockdep hash:
489 */
490 lockdep_init();
73839c5b 491 smp_setup_processor_id();
3ac7fe5a 492 debug_objects_early_init();
42059429
IM
493
494 /*
495 * Set up the the initial canary ASAP:
496 */
497 boot_init_stack_canary();
498
ddbcc7e8 499 cgroup_init_early();
fbb9ce95
IM
500
501 local_irq_disable();
2ce802f6 502 early_boot_irqs_disabled = true;
fbb9ce95 503
1da177e4
LT
504/*
505 * Interrupts are still disabled. Do necessary setups, then
506 * enable them
507 */
44fd2299 508 boot_cpu_init();
1da177e4 509 page_address_init();
ea676e84 510 pr_notice("%s", linux_banner);
1da177e4 511 setup_arch(&command_line);
cf475ad2 512 mm_init_owner(&init_mm, &init_task);
6345d24d 513 mm_init_cpumask(&init_mm);
30d7e0d4 514 setup_command_line(command_line);
e0982e90 515 setup_nr_cpu_ids();
d6647bdf 516 setup_per_cpu_areas();
44fd2299 517 smp_prepare_boot_cpu(); /* arch-specific boot-cpu hooks */
1da177e4 518
9adb62a5 519 build_all_zonelists(NULL, NULL);
83b519e8
PE
520 page_alloc_init();
521
ea676e84 522 pr_notice("Kernel command line: %s\n", boot_command_line);
83b519e8
PE
523 parse_early_param();
524 parse_args("Booting kernel", static_command_line, __start___param,
525 __stop___param - __start___param,
ae82fdb1 526 -1, -1, &unknown_bootoption);
97ce2c88
JF
527
528 jump_label_init();
529
83b519e8
PE
530 /*
531 * These use large bootmem allocations and must precede
532 * kmem_cache_init()
533 */
162a7e75 534 setup_log_buf(0);
83b519e8 535 pidhash_init();
83b519e8
PE
536 vfs_caches_init_early();
537 sort_main_extable();
538 trap_init();
444f478f 539 mm_init();
de03c72c 540
1da177e4
LT
541 /*
542 * Set up the scheduler prior starting any interrupts (such as the
543 * timer interrupt). Full topology setup happens at smp_init()
544 * time - but meanwhile we still have a functioning scheduler.
545 */
546 sched_init();
547 /*
548 * Disable preemption - early bootup scheduling is extremely
549 * fragile until we cpu_idle() for the first time.
550 */
551 preempt_disable();
f91eb62f 552 if (WARN(!irqs_disabled(), "Interrupts were enabled *very* early, fixing it\n"))
c4a68306 553 local_irq_disable();
9f58a205 554 idr_init_cache();
1da177e4 555 rcu_init();
d1e43fa5 556 tick_nohz_init();
65f382fd 557 context_tracking_init();
773e3eb7 558 radix_tree_init();
0b8f1efa
YL
559 /* init some links before init_ISA_irqs() */
560 early_irq_init();
1da177e4 561 init_IRQ();
ad2b1353 562 tick_init();
1da177e4 563 init_timers();
c0a31329 564 hrtimers_init();
1da177e4 565 softirq_init();
ad596171 566 timekeeping_init();
88fecaa2 567 time_init();
38ff87f7 568 sched_clock_postinit();
9e630205 569 perf_event_init();
93e02814 570 profile_init();
d8ad7d11 571 call_function_init();
f91eb62f 572 WARN(!irqs_disabled(), "Interrupts were enabled early\n");
2ce802f6 573 early_boot_irqs_disabled = false;
93e02814 574 local_irq_enable();
dcce284a 575
7e85ee0c 576 kmem_cache_init_late();
1da177e4
LT
577
578 /*
579 * HACK ALERT! This is early. We're enabling the console before
580 * we've done PCI setups etc, and console_init() must be aware of
581 * this. But we do want output early, in case something goes wrong.
582 */
583 console_init();
584 if (panic_later)
585 panic(panic_later, panic_param);
fbb9ce95
IM
586
587 lockdep_info();
588
9a11b49a
IM
589 /*
590 * Need to run this when irqs are enabled, because it wants
591 * to self-test [hard/soft]-irqs on/off lock inversion bugs
592 * too:
593 */
594 locking_selftest();
595
1da177e4
LT
596#ifdef CONFIG_BLK_DEV_INITRD
597 if (initrd_start && !initrd_below_start_ok &&
bd673c7c 598 page_to_pfn(virt_to_page((void *)initrd_start)) < min_low_pfn) {
ea676e84 599 pr_crit("initrd overwritten (0x%08lx < 0x%08lx) - disabling it.\n",
bd673c7c
GU
600 page_to_pfn(virt_to_page((void *)initrd_start)),
601 min_low_pfn);
1da177e4
LT
602 initrd_start = 0;
603 }
604#endif
94b6da5a 605 page_cgroup_init();
3ac7fe5a 606 debug_objects_mem_init();
9b090f2d 607 kmemleak_init();
e7c8d5c9 608 setup_per_cpu_pageset();
1da177e4
LT
609 numa_policy_init();
610 if (late_time_init)
611 late_time_init();
fa84e9ee 612 sched_clock_init();
1da177e4
LT
613 calibrate_delay();
614 pidmap_init();
1da177e4 615 anon_vma_init();
11520e5e 616#ifdef CONFIG_X86
83e68189 617 if (efi_enabled(EFI_RUNTIME_SERVICES))
11520e5e
LT
618 efi_enter_virtual_mode();
619#endif
8c9843e5 620 thread_info_cache_init();
d84f4f99 621 cred_init();
4481374c 622 fork_init(totalram_pages);
1da177e4
LT
623 proc_caches_init();
624 buffer_init();
1da177e4
LT
625 key_init();
626 security_init();
0b4b3827 627 dbg_late_init();
4481374c 628 vfs_caches_init(totalram_pages);
1da177e4
LT
629 signals_init();
630 /* rootfs populating might need page-writeback */
631 page_writeback_init();
632#ifdef CONFIG_PROC_FS
633 proc_root_init();
634#endif
ddbcc7e8 635 cgroup_init();
1da177e4 636 cpuset_init();
c757249a 637 taskstats_init_early();
ca74e92b 638 delayacct_init();
1da177e4
LT
639
640 check_bugs();
641
642 acpi_early_init(); /* before LAPIC and SMP init */
6ae6996a 643 sfi_init_late();
1da177e4 644
83e68189 645 if (efi_enabled(EFI_RUNTIME_SERVICES)) {
2223af38 646 efi_late_init();
78510792 647 efi_free_boot_services();
2223af38 648 }
78510792 649
68bf21aa
SR
650 ftrace_init();
651
1da177e4
LT
652 /* Do the rest non-__init'ed, we're now alive */
653 rest_init();
654}
655
b99b87f7
PO
656/* Call all constructor functions linked into the kernel. */
657static void __init do_ctors(void)
658{
659#ifdef CONFIG_CONSTRUCTORS
196a15b4 660 ctor_fn_t *fn = (ctor_fn_t *) __ctors_start;
b99b87f7 661
196a15b4
HS
662 for (; fn < (ctor_fn_t *) __ctors_end; fn++)
663 (*fn)();
b99b87f7
PO
664#endif
665}
666
2329abfa 667bool initcall_debug;
d0ea3d7d 668core_param(initcall_debug, initcall_debug, bool, 0644);
1da177e4 669
e4461271 670static int __init_or_module do_one_initcall_debug(initcall_t fn)
1da177e4 671{
74239072 672 ktime_t calltime, delta, rettime;
30dbb20e
AW
673 unsigned long long duration;
674 int ret;
1da177e4 675
ea676e84 676 pr_debug("calling %pF @ %i\n", fn, task_pid_nr(current));
22c5c03b 677 calltime = ktime_get();
30dbb20e 678 ret = fn();
22c5c03b
KW
679 rettime = ktime_get();
680 delta = ktime_sub(rettime, calltime);
681 duration = (unsigned long long) ktime_to_ns(delta) >> 10;
ea676e84
AM
682 pr_debug("initcall %pF returned %d after %lld usecs\n",
683 fn, ret, duration);
1da177e4 684
22c5c03b
KW
685 return ret;
686}
687
e4461271 688int __init_or_module do_one_initcall(initcall_t fn)
22c5c03b
KW
689{
690 int count = preempt_count();
691 int ret;
ff1c8fac 692 char msgbuf[64];
22c5c03b
KW
693
694 if (initcall_debug)
695 ret = do_one_initcall_debug(fn);
696 else
697 ret = fn();
8f0c45cd 698
e0df154f 699 msgbuf[0] = 0;
e662e1cf 700
e0df154f 701 if (preempt_count() != count) {
bf5d770b 702 sprintf(msgbuf, "preemption imbalance ");
4a2b4b22 703 preempt_count_set(count);
1da177e4 704 }
e0df154f 705 if (irqs_disabled()) {
a76bfd0d 706 strlcat(msgbuf, "disabled interrupts ", sizeof(msgbuf));
e0df154f
LT
707 local_irq_enable();
708 }
f91eb62f 709 WARN(msgbuf[0], "initcall %pF returned with %s\n", fn, msgbuf);
59f9415f 710
30dbb20e 711 return ret;
e0df154f
LT
712}
713
714
026cee00
PM
715extern initcall_t __initcall_start[];
716extern initcall_t __initcall0_start[];
717extern initcall_t __initcall1_start[];
718extern initcall_t __initcall2_start[];
719extern initcall_t __initcall3_start[];
720extern initcall_t __initcall4_start[];
721extern initcall_t __initcall5_start[];
722extern initcall_t __initcall6_start[];
723extern initcall_t __initcall7_start[];
724extern initcall_t __initcall_end[];
725
726static initcall_t *initcall_levels[] __initdata = {
727 __initcall0_start,
728 __initcall1_start,
729 __initcall2_start,
730 __initcall3_start,
731 __initcall4_start,
732 __initcall5_start,
733 __initcall6_start,
734 __initcall7_start,
735 __initcall_end,
736};
737
96263d28 738/* Keep these in sync with initcalls in include/linux/init.h */
026cee00 739static char *initcall_level_names[] __initdata = {
9fb48c74
JC
740 "early",
741 "core",
742 "postcore",
743 "arch",
744 "subsys",
745 "fs",
746 "device",
747 "late",
026cee00
PM
748};
749
026cee00 750static void __init do_initcall_level(int level)
e0df154f 751{
026cee00 752 extern const struct kernel_param __start___param[], __stop___param[];
196a15b4 753 initcall_t *fn;
e0df154f 754
08746a65 755 strcpy(initcall_command_line, saved_command_line);
026cee00 756 parse_args(initcall_level_names[level],
08746a65 757 initcall_command_line, __start___param,
026cee00
PM
758 __stop___param - __start___param,
759 level, level,
eb157427 760 &repair_env_string);
026cee00
PM
761
762 for (fn = initcall_levels[level]; fn < initcall_levels[level+1]; fn++)
196a15b4 763 do_one_initcall(*fn);
1da177e4
LT
764}
765
026cee00
PM
766static void __init do_initcalls(void)
767{
768 int level;
769
19efb72f 770 for (level = 0; level < ARRAY_SIZE(initcall_levels) - 1; level++)
026cee00
PM
771 do_initcall_level(level);
772}
773
1da177e4
LT
774/*
775 * Ok, the machine is now initialized. None of the devices
776 * have been touched yet, but the CPU subsystem is up and
777 * running, and memory and process management works.
778 *
779 * Now we can finally start doing some real work..
780 */
781static void __init do_basic_setup(void)
782{
759ee091 783 cpuset_init_smp();
1da177e4 784 usermodehelper_init();
41ffe5d5 785 shmem_init();
1da177e4 786 driver_init();
b04c3afb 787 init_irq_proc();
b99b87f7 788 do_ctors();
d5767c53 789 usermodehelper_enable();
b0f84374 790 do_initcalls();
47d06e53 791 random_int_secret_init();
1da177e4
LT
792}
793
7babe8db 794static void __init do_pre_smp_initcalls(void)
c2147a50 795{
196a15b4 796 initcall_t *fn;
c2147a50 797
026cee00 798 for (fn = __initcall_start; fn < __initcall0_start; fn++)
196a15b4 799 do_one_initcall(*fn);
c2147a50
EGM
800}
801
bb813f4c
TH
802/*
803 * This function requests modules which should be loaded by default and is
804 * called twice right after initrd is mounted and right before init is
805 * exec'd. If such modules are on either initrd or rootfs, they will be
806 * loaded before control is passed to userland.
807 */
808void __init load_default_modules(void)
809{
810 load_default_elevator_module();
811}
812
a74fb73c 813static int run_init_process(const char *init_filename)
1da177e4
LT
814{
815 argv_init[0] = init_filename;
ae903caa
AV
816 return do_execve(init_filename,
817 (const char __user *const __user *)argv_init,
818 (const char __user *const __user *)envp_init);
1da177e4
LT
819}
820
ba24762b
MO
821static int try_to_run_init_process(const char *init_filename)
822{
823 int ret;
824
825 ret = run_init_process(init_filename);
826
827 if (ret && ret != -ENOENT) {
828 pr_err("Starting init: %s exists but couldn't execute it (error %d)\n",
829 init_filename, ret);
830 }
831
832 return ret;
833}
834
f80b0c90 835static noinline void __init kernel_init_freeable(void);
d6b21238
AV
836
837static int __ref kernel_init(void *unused)
ee5bfa64 838{
ba24762b
MO
839 int ret;
840
d6b21238 841 kernel_init_freeable();
22a9d645
AV
842 /* need to finish all async __init code before freeing the memory */
843 async_synchronize_full();
ee5bfa64 844 free_initmem();
ee5bfa64
VG
845 mark_rodata_ro();
846 system_state = SYSTEM_RUNNING;
847 numa_default_policy();
848
4a9d4b02 849 flush_delayed_fput();
fae5fa44 850
ee5bfa64 851 if (ramdisk_execute_command) {
ba24762b
MO
852 ret = run_init_process(ramdisk_execute_command);
853 if (!ret)
a74fb73c 854 return 0;
ba24762b
MO
855 pr_err("Failed to execute %s (error %d)\n",
856 ramdisk_execute_command, ret);
ee5bfa64
VG
857 }
858
859 /*
860 * We try each of these until one succeeds.
861 *
862 * The Bourne shell can be used instead of init if we are
863 * trying to recover a really broken machine.
864 */
865 if (execute_command) {
ba24762b
MO
866 ret = run_init_process(execute_command);
867 if (!ret)
a74fb73c 868 return 0;
ba24762b
MO
869 pr_err("Failed to execute %s (error %d). Attempting defaults...\n",
870 execute_command, ret);
ee5bfa64 871 }
ba24762b
MO
872 if (!try_to_run_init_process("/sbin/init") ||
873 !try_to_run_init_process("/etc/init") ||
874 !try_to_run_init_process("/bin/init") ||
875 !try_to_run_init_process("/bin/sh"))
a74fb73c 876 return 0;
ee5bfa64 877
ba24762b 878 panic("No working init found. Try passing init= option to kernel. "
9a85b8d6 879 "See Linux Documentation/init.txt for guidance.");
ee5bfa64
VG
880}
881
f80b0c90 882static noinline void __init kernel_init_freeable(void)
1da177e4 883{
b433c3d4
PZ
884 /*
885 * Wait until kthreadd is all set-up.
886 */
887 wait_for_completion(&kthreadd_done);
31a67102
LT
888
889 /* Now the scheduler is fully set up and can do blocking allocations */
890 gfp_allowed_mask = __GFP_BITS_MASK;
891
58568d2a
MX
892 /*
893 * init can allocate pages on any node
894 */
3c466d46 895 set_mems_allowed(node_states[N_MEMORY]);
1da177e4
LT
896 /*
897 * init can run on any cpu.
898 */
1a2142af 899 set_cpus_allowed_ptr(current, cpu_all_mask);
1da177e4 900
9ec52099
CLG
901 cad_pid = task_pid(current);
902
ca74a6f8 903 smp_prepare_cpus(setup_max_cpus);
1da177e4
LT
904
905 do_pre_smp_initcalls();
004417a6 906 lockup_detector_init();
1da177e4 907
1da177e4
LT
908 smp_init();
909 sched_init_smp();
910
1da177e4
LT
911 do_basic_setup();
912
2bd3a997
EB
913 /* Open the /dev/console on the rootfs, this should never fail */
914 if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0)
ea676e84 915 pr_err("Warning: unable to open an initial console.\n");
2bd3a997
EB
916
917 (void) sys_dup(0);
918 (void) sys_dup(0);
1da177e4
LT
919 /*
920 * check if there is an early userspace init. If yes, let it do all
921 * the work
922 */
ffdfc409
OJ
923
924 if (!ramdisk_execute_command)
925 ramdisk_execute_command = "/init";
926
927 if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) {
928 ramdisk_execute_command = NULL;
1da177e4 929 prepare_namespace();
ffdfc409 930 }
1da177e4
LT
931
932 /*
933 * Ok, we have completed the initial bootup, and
934 * we're essentially up and running. Get rid of the
935 * initmem segments and start the user-mode stuff..
936 */
bb813f4c
TH
937
938 /* rootfs is available now, try loading default modules */
939 load_default_modules();
1da177e4 940}